dsh-task-splitter：在宿主侧识别长任务并自动拆分，适配小上下文本地模型
dsh-task-splitter 是 Passenger-Qin 维护的 DSH 工作流插件。它在宿主侧用启发式识别「长产出」与「多步流程」两类任务：车道 A 搭建 parts 骨架与 MANIFEST，配合 dsh-task-parts 分块写出后 assemble；车道 B 搭建 PLAN.md 与 notes，强调先计划、每轮一步、结果落盘。规程经 agent/pre-step 同请求注入。安装命令为 dsh plugin --profile 添加 dsh-task-splitter，可配置 lanes、minScore、workflowScore 等。适合 llama.cpp/Ollama 小上下文场景；车道 B 保证过程诚实而非答案正确。MIT，Node >= 20。
Read MoreKeenable for DeepSeek Harness：无 Key 的网页搜索与抓取插件
`keenableai/dsh-keenable` 插件为 DeepSeek Harness 提供无需账号的网页搜索与抓取能力。它将 DSH 的内置工具桥接到 Keenable 服务，支持 Markdown 格式返回结果，并允许通过 API Key 提升速率限制。安装后无需配置即可使用，适合本地模型场景，但需注意隐私及非本地地址的限制。
Read Moredsh-connect-trae-plus：本地 Trae 模型接入与加固分支
`dsh-connect-trae-plus` 是 DeepSeek Harness (DSH) 的加固分支插件，用于将本地 Trae 模型接入工作流。它修复了上游插件因环境异常导致的启动失败问题，支持国内与国际版并行接入。该插件提供模型管理、只读用量查询、本地工具循环调用及安全隔离机制，但目前仅支持 Trae SOLO 通道的模型。安装后需重启 DSH 进程。
Read MoreDSH-LocalModels-TokenSavior: Zero Cloud Token Costs for Calling Local Ollama Models
DSH-LocalModels-TokenSavior is a plugin that offloads read-only collection tasks to local Ollama models, aiming to process large files or logs at zero cost. Installation requires Ollama access, Provider routing configuration, and model declaration as prerequisites. It can be enabled through the settings panel or configuration file, and supports checking status with `ollama_local_models()` as well as executing extraction, classification, and other tasks with `subagent_local()`. The plugin returns only conclusions, does not consume the main conversation context, and is suitable for use cases such as log analysis.
Read MoreContext Guardian: A Practical Solution for Local Model Context Management
Context Guardian is an open-source plugin maintained by LuminariSoftwares, designed to resolve session interruptions caused by context window overflow in local models. It offers two modes—an OpenAI-compatible proxy and a DeepSeek Harness native engine—with support for deterministic compression. Its key features include fail-open behavior, disk persistence of raw data, and retrieval of archived context. Users can enable this functionality by configuring the proxy or integrating DSH, ensuring stable sessions and preserving historical details while keeping context length within limits.
Read Moredsh-wsl-ollama: Connect to Local Ollama Models on WSL
DSH Harness (DSH) provides the `dsh-wsl-ollama` plugin to resolve network connectivity issues when accessing local Ollama in a WSL environment. This optional plugin supports status queries, model listings, local chat, and vectorization. It is installed through a specific command, and its configuration supports automatic host address detection or manual setup. It is released under the MIT license.
Read MoreIntroducing the DSH MiMo Connect Plugin
The DSH MiMo Connect plugin is designed to integrate models from the MiMo desktop client into DeepSeek Harness with zero configuration. It supports reusing the client's login credentials and quota without requiring an additional API key, and features lightweight operation, image input support, and streaming chain-of-thought display. After installation, users can directly select and use models under the MiMo group in DSH. The plugin does not depend on the client being running and supports local login verification, but note that the free quota has been discontinued and an account subscription is required.
Read Moreiyicode/dsh-ctx-probe: Automatically Detect Local Model Context
The dsh-ctx-probe plugin resolves 400 overflow issues caused by configuration mismatches in the DSH llm-pi-ai local server. It probes the actual context windows of llama.cpp and Ollama servers in real time, enabling bidirectional automatic synchronization with local settings. To prevent probe storms, it uses single-flight and exponential backoff strategies, and supports real-time probing for local requests and caching for remote requests. After installation, restart DSH Web Host; use the `/ctx-probe force` command to view the configuration, effective settings, and probed values.
Read MoreRunning Local Models in DeepSeek Harness: vmarcelo49/dsh-local-models
DeepSeek Harness (DSH) introduces a new `dsh-local-models` plugin that enables direct invocation of local `.gguf` models within the web interface. Designed to work out of the box, the plugin provides model selection, context and KV cache tuning, real-time VRAM estimation, configuration persistence, and OpenAI-compatible routing. After installation, users must restart DSH to load models through the interface and register them as LLM providers, with support for automatic startup and idle eviction.
Read MoreIntroducing the robbywang25/dsh-llm-mlx Plugin
The `robbywang25/dsh-llm-mlx` plugin is a local model adapter for DeepSeek Harness (DSH), designed specifically for Apple Silicon macOS. It connects local MLX-LM or MLX-VLM models to DSH via an OpenAI-compatible interface, offering two operating modes: hosted mode (where DSH manages the process) and reuse mode (connecting to an existing server). The plugin must be used with a specific version of DSH, requires Python and the corresponding MLX libraries to be installed, and does not include model weight files.
Read Moredsh-llamacpp: Local llama.cpp Adapter
`dsh-llamacpp` is an adapter plugin designed specifically for DeepSeek Harness (DSH) to connect to a local llama.cpp server. It communicates through an OpenAI-compatible interface, registers as the `llamacpp` provider in Harness, and supports streaming output and text-based tool calling. The plugin allows the server address, model name, and API key to be configured via environment variables or a configuration file. It currently supports text only, does not support images or Ollama, and includes robust error handling.
Read MoreDSH Loomy Connect: Zero-Configuration Integration with Loomy's Built-in Models
DSH Loomy Connect is a DeepSeek Harness plugin designed to solve the complexity of configuring local model integrations for developers. By reading the login state and model list of the Loomy desktop application in read-only mode, it registers them as a DSH model provider to enable zero-configuration access. It supports automatic model discovery, displays account and credit information, and provides bilingual support in Chinese and English. Installation requires the Loomy desktop application and DSH version >= 0.1.5-rc.1. The plugin only performs local file reading and authentication, with clear security boundaries, and is suitable for personal learning and research.
Read Moredsh-context-budget: A Plugin for Keeping Local Model Context Within a Controllable GPU Range
The `dsh-context-budget` plugin is designed to address the sharp performance degradation of local models under long contexts. It measures prefill speed and enforces hard limits and checks for token count, first-token latency, and cold prefill time. When a limit is exceeded, it can print a warning or perform compaction, keeping the context within the range the GPU can process. The plugin coexists with `compaction-basic` and provides a `/context-budget` command to monitor status, helping users optimize the experience of using local large language models.
Read MoreDeepSeek Harness Network Capability Enhancement Plugin: dsh-webstack
dsh-webstack is a web access plugin for DeepSeek Harness, designed to address the limited context windows of local models and anti-scraping challenges. It integrates SearXNG and Scrapling to provide a multi-layer scraping pipeline that automatically handles static pages, dynamic rendering, and advanced anti-scraping measures such as Cloudflare. It achieves 73.1% token savings through Turndown conversion and includes strict SSRF security safeguards. Developers can invoke it seamlessly simply by configuring a Provider, without modifying any code.
Read MoreIntroduction to DeepSeek Harness Local Model Plugin
`dsh-plugin-local-model` is a plugin for DeepSeek Harness designed to simplify local model inference. It uses a dedicated settings page to automatically launch llama.cpp upon the first conversation message after a model has been configured, and it automatically unloads after 5 minutes of inactivity to release resources. The plugin also supports one-click switching between parameter presets, ensuring offline operation and file security. Users must prepare their own models and tools; for installation, it is recommended to refer to the source code. This tool is suitable for developers who need local deployment, privacy protection, and frequent parameter switching.
Read Moremodelspoke: DeepSeek Harness Local Model Discovery and Metadata Parsing Plugin
modelspoke is a local model server discovery and metadata parsing plugin specifically designed for DeepSeek Harness (DSH), intended to simplify cumbersome manual configuration and enable advanced features. It supports servers such as llama-swap and Ollama, providing discovery and configuration of common model presets, reasoning effort levels, and image input capabilities through an easy-to-use UI. Installation requires adding it via the pnpm command; once enabled, the server address, API key, and model details can be configured in the sidebar. The plugin runs with the current process permissions, so source code security should be considered, and the model list is curated.
Read Moredsh-fetch-timeouts: Increase HTTP Timeout Settings for DeepSeek Harness
The `dsh-fetch-timeouts` plugin is designed to address connection interruptions in DeepSeek Harness (DSH) during local model inference caused by Node.js's default 5-minute timeout mechanism. By raising Node.js's global HTTP timeout threshold to 30 minutes, it allows requests involving prolonged periods without a response or data transfer (such as large file generation) to complete successfully. After installation, `headersTimeoutMs` and `bodyTimeoutMs` must be configured, and watchdogs must be set in provider routes accordingly. The plugin takes effect globally and affects all operations using Node's global fetch, so it should be evaluated carefully.
Read Moredsh-plugin-grok: DeepSeek Harness Plugin for Locally Driving the Grok Build CLI
`dsh-plugin-grok` is a developer preview plugin for DeepSeek Harness (DSH) that drives the local Grok Build CLI as a tool, providing models with text, image, and video generation capabilities. Instead of relying on HTTP calls, the plugin directly executes the local `grok` binary and supports authentication via `grok login` or `XAI_API_KEY`. Usage requires proxy configuration and interacts with the three exposed tools: `grok_ask`, `grok_imagine_image`, and `grok_imagine_video`. Video generation works by first producing a static image and then converting it into a video. The plugin is currently released as a developer preview under the MIT license.
Read MoreDeepSeek Harness Memory Tree Bridge Plugin
`dsh-memory-bridge` is a memory bridge plugin for DeepSeek Harness (DSH), designed to address the limitations of local model context windows (4K–32K). By distilling conversations into a visualized memory tree and providing persistent storage and retrieval, it enables agents to have long-term memory. The plugin’s core capabilities include automatic memory card accumulation, LLM-free retrieval (BM25 + RRF), memory governance (forgetting curves and auditing), and profile distillation. Its architecture follows a Host JS + Python Sidecar pattern, supports plaintext Markdown storage, and ensures data controllability and fault isolation. Through a dual-channel write strategy, it effectively resolves the pain points of agent “forgetfulness” and “context overflow.”
Read MoreDeepSeek Harness Plugin: dsh-llama-responses
The DeepSeek Harness plugin dsh-llama-responses adopts a plugin-based architecture that routes sub-agent requests to a local llama.cpp service over the OpenAI protocol, reducing costs and protecting privacy. It supports multi-turn memory, streaming tool calls, and sub-agent isolation. During installation, make sure llama.cpp is running; once the adapter is configured, sub-agents can use a local model (e.g., qwen3.8-27b). Enabling the sub-agent tool in the configuration file allows the main agent and local sub-agents to operate independently.
Read MoreDeepSeek Harness Plugin: doubao-dsh-plugin
The doubao-dsh-plugin for DeepSeek Harness (DSH) enables seamless invocation of Doubao model capabilities in local model development. The plugin relies on a locally running Doubao Relay in Quicker and, by injecting the `doubao_ask` tool, supports web search, image generation, multimodal image recognition, and “paste image to path” functionality. Installation requires configuring the Relay first, then adding the plugin via npm and restarting Harness. The plugin supports multiple input formats, automatically handles image compression and deduplication, and is suitable for direct command interaction within conversations.
Read MoreDeepSeek Harness Enhancement Plugin: dsh-plus
nanbbb/dsh-plus is an enhanced plugin for DeepSeek Harness (DSH), designed to streamline workflows through a Web interface. The plugin integrates four core features into the settings panel: vision and memory configuration, a plugin marketplace (supporting browsing and installation), local model management (supporting Ollama/LM Studio), and one-click engine restart. Install it using `dsh plugin add github:nanbbb/dsh-plus`; desktop users do not need to install it separately.
Read MoreDeepSeek Harness Plugin: dsh-opencode-provider
dsh-opencode-provider is a plugin for DeepSeek Harness (DSH) designed to connect DSH to a local OpenCode server over HTTP, enabling local model invocation without an API key. It supports streaming output, retry policies, and event compatibility, but is limited to text-based input and output. It does not support structured tool calling or multimodal capabilities, and requires the local server to operate within a restricted network environment. This plugin offers developers a convenient, low-cost entry point for local inference.
Read Moredsh-unsloth-hands: Offloading repetitive tasks of DeepSeek Harness online models to local Unsloth Desktop
dsh-unsloth-hands is a pure client-side plugin maintained by MicroHEROX (MIT licensed, v0.1.0), allowing DeepSeek Harness online models to perform repetitive short tasks such as batch rewriting, translation, string cleaning, and screenshot OCR.
Read More